Description
Anticipate sociotechnical failure by walking checklists of cognitive, communication and social pathologies over task, discourse and power-trust models, turning each risk into a requirement.
Sutcliffe's method models a work system at three levels and applies a checklist of known pathologies to each. Work is described with scenarios and reusable generic task templates; cognitive pathologies, organised by skill, rule and knowledge levels, are checked against each task and countermeasures are added as requirements. Communication is modelled with discourse patterns and checked for communication pathologies such as missing acknowledgement. Finally, power and trust relationships between agents are modelled in an i*-style dependency view and checked for social pathologies. The same models are then used to compare alternative designs, for example different allocations of roles between people and automation. Its strength is breadth: it brings human error, communication breakdown and organisational politics into one requirements exercise, though the pathology lists are judgement aids, not validated predictors.
Attributes
- Discipline
- Human Factors & Error
- Loss type
- Latent
- Stage
- Definition, Design
- Data
- Analytical / expert
- Effort
- Weeks
Steps
- Model the work with scenarios and generic task templates.
- Apply the cognitive pathology checklist (skill, rule and knowledge levels) to each task and record countermeasures as requirements.
- Model communication between agents with discourse patterns and apply the communication pathology checklist.
- Model power and trust dependencies between agents and apply the social pathology checklist.
- Attach failure-point warnings and countermeasures to the requirements specification.
- Use the models to compare alternative designs or role allocations.
Sources
- Handbook of Cognitive Task Design (Hollnagel, 2003)
Also known as
- Sociotechnical pathology checklist analysis
- Generic tasks, discourse patterns and power-trust analysis
Categories
- Accessibility
- Human Factors & Error